Switch-Mode Power Supply

By Peter Lu, Product Engineer at GreenMore | September 20, 2026 This is the second article in our four-part DC power guide. Part 1 covered how to choose a DC power module for an energy storage cabinet; here we look under the hood at the two numbers engineers argue about most: conversion efficiency and output ripple. Inside every modern cabinet battery converter, transistors switch thousands of times per second, chopping one DC voltage and reshaping it into another. The GreenMore GM-LDC30 and GM-LDC60 use industrial-grade high-frequency conversion to reach peak efficiencies of 98.0% and 98.6%. That sounds like a small gap. Across a 60 kW power stage running day and night, it is not. How Does High-Frequency DC-DC Conversion Work, and Why Is It Efficient? A switch-mode DC-DC converter does not burn off excess voltage the way a linear regulator does. It stores energy briefly in an inductor and transfers it in packets timed by the switch. The ratio of on-time to off-time sets the output voltage. Because the transistor is either fully on or fully off, the average dissipation in the switch stays low. Raise the switching frequency and the passive components shrink. As Texas Instruments explains, the required output inductance and capacitance are both inversely proportional to switching frequency. Double the frequency and you need roughly half the inductance, so the magnetic core gets smaller, lighter, and cheaper. That is the core reason a 60 kW converter fits in a rack-mount enclosure today while needing a cabinet of its own a generation ago. Efficiency is never free, though. At high frequency the loss budget shifts between three main components: Conduction loss I²R heating in switch on-resistance, inductor windings, and busbars Independent of frequency; grows with load Switching loss Energy lost while voltage and current overlap during each turn-on and turn-off Increases directly with frequency Core and drive loss Magnetic hysteresis in the core and energy needed to charge the gate each cycle Increases with frequency Resonant and phase-shifted control schemes soften the voltage-current overlap by timing the switch transition to a zero crossing, cutting the switching term. The designer's job is to pick the frequency, topology, and silicon that balance these terms. At heavy load, conduction loss dominates, so low on-resistance devices win. At light load, switching and core losses dominate, which is why the module draws less than 20 W in standby rather than idling at full magnetizing current. Schematic diagram of low ripple power supply technology What Do 98.0% and 98.6% Efficiency Actually Buy You? Efficiency converts directly into heat, which is what makes a high efficiency DC power supply module worth its price. Take each module at its rated output: Module Rated Output Peak Efficiency Power That Becomes Heat GM-LDC30 30 kW 98.0% ~612 W GM-LDC60 60 kW 98.6% ~852 W Both figures are stated at the favorable operating point; real efficiency varies with voltage ratio and load. Even so, the relationship is fixed. Every 0.1% of efficiency lost at 60 kW adds 60 W of heat, and that heat has to leave the module or it raises component temperatures. Smart air cooling handles it across the full −30 to +55 °C operating range, but hotter capacitors and semiconductors age faster. TDK notes that capacitor life is governed by thermal stress, with heat from ripple and conduction losses the deciding factor. The energy meter notices too. At a common 60 kW output, the gap between 98.0% and 98.6% efficiency is about 373 W of dissipation, or roughly 3,260 kWh per module per year if it ran continuously. Few cabinets run flat-out all year, but sites with long cycling duty see the number on their electricity bill. That is why the standby figure matters. Under 20 W keeps parasitic losses negligible when the cabinet is parked, instead of paying for a converter that wastes more in a day idle than it should. Why Low Ripple Matters for Batteries and Downstream Equipment Ripple is the periodic AC component riding on the DC output, at the switching frequency and its harmonics. It is distinct from random high-frequency noise, though the two get measured together on a scope. The output filter exists to deliver a low ripple DC output that every load on the bus can trust. Batteries are the first victim people cite. A Solis white paper reviewing 16 kW inverter testing reported that 6 A peak-to-peak ripple near 300 Hz accelerated capacity fade by up to 15% compared with smooth DC cycling, driven by localized heating and uneven lithium intercalation (Solis, The Role of Ripple Current on Lithium Battery's Lifecycle). The effect is not universally settled. Long-term academic studies reach different conclusions, and a 2022 battery-aging dataset study in Batteries found ripple effects varied with frequency and cell design. The consistent finding is that low-frequency, high-amplitude ripple does the damage; high-frequency ripple is largely filtered by the cell's double-layer capacitance. No cabinet designer should gamble a battery warranty on that distinction when the conversion stage can simply be clean. Sensitive loads share the same bus: Load How Ripple Hurts It Telecom rectifiers and radios Raises noise floor, can push transmit signal quality out of spec Sensors and analog front ends Small measurement offsets and jitter at the switching frequency Automation controllers and PLCs Logic brownouts and communication errors when dips cross thresholds DC-link capacitors Extra I²R heating shortens service life, the classic wear-out path Ripple is a steady-state power-quality problem, not a fault event. Protection such as OVP, OCP, SCP, and OTP handles abnormal conditions, which is the subject of our protection deep dive. How Should Output Ripple Be Measured? A sloppy probe setup measures the probe's own antenna loop instead of the converter. Industry practice converges on a few rules. Intel's ATX 3.0 power supply design guide defines ripple and noise over 10 Hz to 20 MHz and requires the oscilloscope bandwidth set to 20 MHz, which keeps wideband switching spikes out of the periodic ripple number. On the bench: Put the scope in AC coupling and engage the 20 MHz bandwidth limit for the ripple reading. Use a 1:1 probe or a dedicated power-rail probe for millivolt sensitivity. A 10:1 probe throws away the vertical resolution you need. Connect the probe tip directly across the output capacitor and use the short spring ground clip, not the dangling ground lead. A long ground lead is an antenna. Measure both peak-to-peak and RMS, at no load and full load. When you compare module datasheets, check the conditions first. A ripple number without a bandwidth limit and load point tells you nothing. Where the GM-LDC Modules Fit The GM-LDC30 (30 kW) and GM-LDC60 (60 kW) are bidirectional, rack-mount units. As a DC-DC converter for energy storage duty, each moves power in both directions between battery and bus. The low-voltage side operates from 200 to 900 V and the high-voltage side from 300 to 1000 V, with maximum output current of 75 A for the 30 kW unit and 150 A for the 60 kW unit. Smart air cooling, IP20 ingress protection, RS485 communication, reverse-polarity and surge protection, insulation monitoring, and a three-year warranty come standard, and modules support parallel operation for larger power stages. They sit inside air-cooled and liquid-cooled cabinet systems as well as utility-scale containerized ESS. When your cabinet needs more than 60 kW, our next article covers parallel expansion and current sharing. The GreenMore team has worked in solar and power electronics since 2017. If you are sizing the DC stage for a cabinet or container project and want help matching efficiency and ripple targets to your battery and load, talk to our engineering team.